Stabilization of the frequency difference between a pair of semiconductor lasers is an important goal in the development of optical communication systems. In the past, a number of techniques for achieving this goal have been proposed or attempted. In one prior technique, the output of a laser diode is modulated at a fixed frequency, thus producing multiple sidebands in the laser's power spectrum. The gain bandwidths of two "slave" lasers are individually tuned to overlap different ones of these sidebands, and the slave lasers are then injection locked to the respective sidebands. As a results, the frequency separation between the slave lasers is a multiple of the modulation frequency. A drawback of this technique is that the frequency stability is governed by the drift of the frequency of the master laser. In addition, the linewidths of the slave lasers will tend to be greater than or equal to that of the master laser. This latter phenomenon is due to the action of the sidebands of the master laser's power spectrum outside the slave laser's gain bandwidth, which action introduces noise and broadens the linewidth.
A second proposal for the frequency stabilization of semiconductor lasers is to use an acousto-optic modulator to frequency shift the laser output. The frequency separation between the shifted and unshifted beams is then fixed by the frequency used to drive the modulator. However, in order to achieve multi-gegahertz separations, it may be necessary to employ more than one acousto-optic modulator. In general, this approach is not economically competitive with other techniques for achieving the same end.
In both of the approaches discussed above, the frequency stability and the linewidth of each output beam are governed primarily by the master laser. Although this property can be altered by the addition of an electronic feedback network, the linewidths and/or spectral purity of the first two approaches will still suffer from the effects fo modulation of the master laser, incomplete spectral filtering, and noise from spurious inputs to the slave lasers. The frequency separation in both proposals would be dictated by the frequency of the RF source used to modulated the master laser drive the acousto-optic modulator.
A third prior approach for frequency stabilization is to have a slave laser which is locked to a fixed frequency separation from a master laser via electronic feedback. The electronic feedback system detects the beat signal of the interference of the outputs of the master and slave laser, compares it to a reference frequency, and then adjusts the current supplied to the slave laser to maintain a fixed frequency separation. The frequency stability is again dependent on the stability of the master laser.
All three proposals described above require stabilized RF electronics, along with the required optical components. In addition, a significant limitation of these proposals is that the frequency separation that can be obtained between the lasers is limited by the frequency available from the RF source, or from the relatively low order harmonics of the RF frequency. This requirement significantly limits the usefulness of these techniques in microwave and millimeter-wave applications.